January 27, 2012
Quality ADME Data & Models for Non-Experts to Reduce Animal Use - A Guest Blog by George Grass, Pharm.D. Ph.D at info@ADMEdata.com
Animals as surrogate test subjects for humans are becoming less desirable as the effects of cost, negative social pressure, and the reality that animals are not always accurate predictors of human outcome move research to consider alternatives. As computational power increases, and understanding of chemical and physiological systems improves, novel research relies more and more on predictive models. But these models require data to develop and test datasets to prove their validity.
ADMEdata.com is a unique database of ADME related data. The database was created specifically for the purpose of building the predictive ADME simulation model known as iDEA (in vitro Determination for the Estimation of ADME). Maximizing the predictive capability of the models was the primary goal of the project. Since we knew predictive capability was directly a function of data quality, the creation of unique, high quality data sets was the initial challenge.
Because data collated from publications and other disparate sources leads to datasets with high variance and unknown quality, iDEA began not with a modeling effort, but with the generation of a unique database from carefully designed experiments. The first mandate was to chose assays that were determined to be the most predictive for in vivo ADME, and not rely on whatever data was previously available, or happened fit a specific high throughput assay. Second, for each assay, we generated all of our data using a single rigorously validated protocol, run in our own laboratory. Third, the selection of the compounds and assays were conducted with the goal of building maximally predictive models. Therefore compound diversity was guided not by chemical structure alone, but also by mechanisms of absorption, metabolism and elimination, range of values across all of the assays, and range of in vivo pharmacokinetic performance.
The resulting database was never publically disclosed. It was not available to the consortium that funded the iDEA development, the customers who ultimately purchased the iDEA product, or anyone outside of the development team. Now for the first time, this data (and corresponding models) are available to computational and experimental scientists to create and test hypotheses guided with quality ADME data and models.
In summary, these datasets are not collated from literature or collected from the efforts of a variety of discovery projects. This unique database was created for a singular purpose, the creation of a predictive ADME model and given the current economic environment (in the post-doc.com era) it is unlikely that this level of effort will ever be duplicated for generating such a database.
The combination of these data sets with the quality CDD software developed over the past 8 years provides another dimension of added value. One can mine the ADME data in combination with their own private data, compare structures, perform complex Boolean queries, compare calculated physical chemical with experimental properties (such as solubility across 5 different physiologically relevant pH values). Over time, additional datasets will be added to those currently available on the CDD site.
For full data sets and models, directly contact George Grass, Pharm.D. Ph.D at info@ADMEdata.com (distributed via CDD).
For collaborative research ideas directly with these data sets or inspired by them (either with existing or future grants), please contact Barry Bunin, PhD at info@collaborativedrug.com.
This blog is authored by members of the CDD Vault community. CDD Vault is a hosted drug discovery informatics platform that securely manages both private and external biological and chemical data. It provides core functionality including chemical registration, structure activity relationship, chemical inventory, and electronic lab notebook capabilities!
CDD Vault: Drug Discovery Informatics your whole project team will embrace!